fix(xugu): accept scheduler job object sources

[mapan0424]

Summary

Fixes scheduler job source loading for XuguDB connections.

Why this is needed

The desktop tree already exposes scheduler jobs and maps the selected node to the JOB object type. The XuguDB agent also provides the corresponding source reconstruction for scheduler jobs. However, the dbx-core object-source model did not accept JOB when deserializing the request. As a result, selecting View Source failed before the request reached the agent with an error similar to:

invalid args objectType for command get_object_source: unknown variant JOB

This made an object that was visible and otherwise supported by the XuguDB integration appear unusable in the UI.

Implementation

• Added Job to the shared ObjectSourceKind enum.
• Added JOB keyword parsing and serde round-trip coverage.
• Added exhaustive handling in schema/source-generation and transfer paths.
• Kept non-Xugu providers unchanged: their job branches are no-op/unsupported, and Oracle does not attempt to call an invalid metadata API for jobs.
• Preserved the existing XuguDB agent source-generation path without changing its SQL behavior.

Validation

• cargo check -p dbx-core
• cargo check -p dbx-web -q
• cargo test -p dbx-core --lib object_source_sql::tests:: (70 passed)
• cargo test -p dbx-core --lib types::tests::object_source_kind_accepts_job_wire_value -- --exact (passed)
• go test ./... in agents/drivers/xugu (passed)
• Built the debug macOS application with pnpm tauri build --debug --bundles app --no-sign.
• Performed a packaged UI verification: expanded the scheduler-jobs node, selected a job, and opened its source successfully. The generated DBMS_SCHEDULER.CREATE_JOB(...) source was displayed and the unknown-variant error did not recur.

The full dbx-core library suite was also attempted; it reaches an unrelated existing stack-overflow failure in table_export::tests::external_driver_table_export_closes_cursor_at_row_limit. No changed-code test failed.

[t8y2]

Maintainer conflict resolution pushed in ada7b54 (old head 5500169, merged main@7e24f4007 — merge commit with both sides preserved). The single conflict was the tail of mod tests in crates/dbx-core/src/types.rs: this PR's object_source_kind_accepts_job_wire_value vs main's completion_candidate_kind_accepts_uppercase_agent_wire_values. Resolution: keep BOTH tests. All other files auto-merged or were untouched by main; git range-diff maps the PR commit identically, and every PR-added line (the Job variant, all no-op match arms in schema/transfer/object_source_sql, both wire-value tests) is present in the resolved tree.

Checks: git diff --check clean, no conflict markers; Rust CI on this head is the merge gate (targeted types tests were interrupted locally by a disk-full environment issue, not by the code).

[t8y2]

Thanks for the contribution! Merged in 1818387, will be released in the next version.